Kal

An update on the language evaluations. we’ve finished the first Tigrinya round, and now we’re looking for Sidaamu Afoo, or Wolaytta speakers. If you speak either language and have some time to help with the voice evaluation, you can join the trial. We provide 100 ETB as a small thank-you for your time. After the trial, we’ll keep a few strong evaluators close and work with them regularly on future model iterations and evaluations. You can join through Telegram or Discord: ✅ https://t.me/+INclR6k6G3FhMmM0 💰 https://discord.gg/ajuXM4zFA

We have to start practicing saying please to Claude before Nov 12 😭

5,000 hours of speech in the three months since Voices 2 launched. Thank you for every word 🫶 To celebrate, Addis Voices 2 and Addis Scribe are 50% off until Nov 12. Voices 2: 2.5 ETB per minute (was 5 ETB) Scribe: 1.75 ETB per 1,000 characters (was 3.5 ETB) 😊 addisassistant.com @addisassistantai

For those of you wondering where to get Addis Creatives and how it works, here’s a short walkthrough. @kalBuilds

You know when the video is already done, but the captions are just off enough that someone still has to go through them word by word for an hour? 😭 That was one thing I kept coming back to while we were building Addis Creatives. If you still have to fix the timing after, then what’s the point? @kalBuilds

When you download Addis Creatives on Mac or Windows, it checks if you already have Premiere Pro, After Effects or Creative Cloud installed. If it finds them, you’ll get an option to install Addis Creatives as a panel inside those apps. You can also export the captions and use them in any editor of your choice, like CapCut. @kalBuilds

Auto captions for Amharic and turning meeting recordings into timestamped notes just got a whole lot easier. @kalBuilds

Addis Scribe now returns timed captions. Add

timestamps=word

to a transcription and get word-level timestamps, plus ready-to-use SRT or VTT files for Amharic audio. Available in the API and in SDK 0.5.0 ( 📱 npm i addisai · 📱 pip install addisai). 😊: docs.addisassistant.com/docs/capabilities/speech-to-text#timestamps-and-captions @addisassistantai

voice streaming got even faster. First audio starts arriving in 500–900 ms at p95. Try it in the doc playground @b1n1yamBuilds

Just redeemed our team yearly subscription from Claude for Startups 🎉 If you’re building an AI startup, go get yours. Seriously, apply. @b1n1yamBuilds

This is turning into a really good week for releases. Yesterday Scribe, today voice streaming is live on Voices 2. Audio now starts in about a second while the rest keeps generating. A lot of what we’ve been building around Realtime voice needed this. Have fun with it. @kalBuilds

A small step toward AI that works where people actually are. My favorite part of the day! 🫡 @kalBuilds

Addis Scribe running fully offline on-device. No Wi-Fi or mobile data, and it keeps transcribing locally (first text in ~0.3s) Small, open-source models running on the edge is something we’ll be doing a lot more of. @b1n1yamBuilds

😶

Coming up next 🤗

😶

coming up next 👀

Seeing the words show up while you’re still talking instead of waiting for the recording to finish is such a different experience. Also, it’s open source, so you can grab the weights from Hugging Face, or just try it in the Playground first and see how fast it feels. 😊Playground 😊Hugging Face

Scribe stream is a really fast model. Give it a try on the playground

We used to serve one request at a time. The GPU had room for 24 and we gave it 1. Under traffic only 8% of Amharic requests made it through. Once we let requests share the GPU it went to 99%. Same machine. so that is 24 users using the same model at once @b1n1yamBuilds #inference